Skip to content
Hot eventDeveloping

UC San Diego releases open ORA v0.1 robot-use agent harness

1 reports1 reporting sourcesUpdated 13 h ago

Start here

AI digest

UC San Diego researchers released ORA v0.1, an open-source harness that lets a language model operate a robot through cameras and tools without any robot training. With DeepSeek-V4.1-Flash, ORA reaches 39.9% zero-shot on RoboDojo, up from 5.2% in October's preview. In code mode the model writes Python against the same robot tools, carrying 3.62 robot actions per turn instead of 1.10, and collision-stopped moves on RoboDojo fall from 14.1% to 7.0%. On RoboLab-120, Qwen3.8-27B goes from 5.4% alone to 52.5% inside ORA, and to 61.9% with PEAR's training added. ORA is open source; running it requires going to the repository.

Generated from reports · updated 13 h ago

Report timeline

Follow the reports to see the event from different sides.

November 19
  1. Hacker News · AI
    ORA v0.1: Open robot-use agent harness reaches 39.9% zero-shot on RoboDojo with DeepSeek-V4.1-Flash

    UC San Diego researchers released ORA v0.1, an open harness in which a language model drives a robot through cameras and tools without robot training, reaching 39.9% zero-shot on RoboDojo with DeepSeek-V4.1-Flash, up from 5.2% in October's preview. In code mode the model writes Python against the same robot tools, carrying 3.62 robot actions per turn instead of 1.10, and collision-stopped moves fall from 14.1% to 7.0% on RoboDojo. On RoboLab-120, Qwen3.8-27B goes from 5.4% alone to 52.5% inside ORA and 61.9% with PEAR's training added.

Heat for this event

Current, comparable range
7
Peak, comparable range
10November 19 15:00
Change in 24 hours
–

The trend compares the same participants over hours observed continuously, so its range can be narrower than the current heat count. Hover or tap the chart for hourly heat; the arrow keys move through it.